Re: Resource management, backend application

On 07/01/2025 14:55, Alex Rousskov wrote:
> On 2025-01-07 04:49, Tony Albers wrote:
>> Is it possible in squid to ensure that a badly behaving backend
>> application doesn't eat up all squid resources?
>
> Yes, especially if you know about that application behavior in advance.
> You can configure Squid to start denying requests for the problematic
> application once the number of concurrent requests for that application
> exceeds some threshold.
>
> You will probably have to use external ACLs to track that concurrency
> level. I do not have a blueprint ready, but it should be doable in
> principle.
>
>
>> E.g.: at work we have an Apache reverse proxy in front of a number of
>> backend hosts. If one of the backend applications misbehaves, this can
>> result in all of apache's worker processes being held up by this
>> application, resulting in apache hanging and all sites going offline.
>> In apache, AFAIK, there is no way to prevent this.
>
> Squid worker processes are not dedicated to a single request or a single
> application so, as Matus UHLAR has already said, the above scenario is
> not going to happen with Squid (but an application can exhaust other
> resources such as socket descriptors or memory, so Squid can be slowed
> down in a similar scenario unless you configure it specially).
>
>
> HTH,
>
> Alex.
>
>
>> But can squid handle this scenario in a way that only the site with the
>> misbehaving application goes offline without pulling the other sites
>> down with it?
>>
>> I understand that the way squid and apache works is different, but
>> that's not really important for me. I just want to use the best tool for
>> the job.
>>
>> TIA,
>>
>> /tony

Thanks Alex and Matus, much appreciated.

/tony